Received: 1 November 2018 | Revised: 25 March 2020 | Accepted: 28 April 2020 DOI: 10.1111/mec.15468

ORIGINAL ARTICLE

The socially parasitic mexicanus has host- associated genetic population structure and related neighbouring colonies

Joseph R. Sapp1 | Jenn Yost2 | Bruce E. Lyon1

1Department of Ecology and Evolutionary Biology, University of California, Santa Cruz, Abstract CA, USA The genetic structure of populations can be both a cause and a consequence of eco- 2 Biological Sciences Department, California logical interactions. For parasites, genetic structure may be a consequence of pref- Polytechnic State University, San Luis Obispo, CA, USA erences for host or of mating behaviour. Conversely, genetic structure can influence where conspecific interactions among parasites lay on a spectrum from Correspondence Joseph R. Sapp, Department of Ecology cooperation to conflict. We used microsatellite loci to characterize the genetic struc- and Evolutionary Biology, University of ture of a population of the socially parasitic dulotic (aka “slave-making”) ant (Polyergus California, Santa Cruz, CA, USA. Email: [email protected] mexicanus), which is known for its host-specificity and conspecific aggression. First, we assessed whether the pattern of host species use by the parasite has influenced Funding information University of California, Santa Cruz parasite population structure. We found that host species use was correlated with Department of Ecology and Evolutionary subpopulation structure, but this correlation was imperfect: some subpopulations Biology; Sigma Xi; University of California Natural Reserve System; University of used one host species nearly exclusively, while others used several. Second, we ex- California, Santa Cruz Services for Transfer amined the viscosity of the parasite population by measuring the relatedness of pairs And Re-entry Students (STARS) of neighbouring parasitic ant colonies at varying distances from each other. Although natural history observations of local dispersal by queens suggested the potential for viscosity, there was no strong correlation between relatedness and distance between colonies. However, 35% of colonies had a closely related neighbouring colony, indi- cating that kinship could potentially affect the nature of some interactions between colonies of this social parasite. Our findings confirm that ecological forces like host species selection can shape the genetic structure of parasite populations, and that such genetic structure has the potential to influence parasite-parasite interactions in social parasites via inclusive fitness.

KEYWORDS dulosis, kin structure, microsatellites, population structure, population viscosity, social

1 | INTRODUCTION (Davies, 2010; Field, 1992; Wisenden, 1999; Zink, 2003). Social parasitism has been particularly well studied in brood parasitic Social parasites exploit the reproductive investment of their hosts for birds (Davies, 2010; Rothstein, 1990) but it is perhaps taken to even their own fitness benefit. Social parasitism occurs in a broad diver- greater extremes among social insects (Aron, Passera, & Keller, 1999; sity of taxa, including birds, fish, and —particularly social Heinze & Keller, 2000; Hölldobler & Wilson, 1990). Most work on

2050 | © 2020 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/mec Molecular Ecology. 2020;29:2050–2062. SAPP et al. | 2051 social parasites has focused on host-parasite dynamics, with far less queen parasitizes an intact colony of a host ant species by killing the attention paid to interactions between the parasites themselves (but resident queen and usurping her workers; and (b) the dulotic queen's see Brooker & Brooker, 1989 for a notable exception for an avian sterile daughters (dulotic workers) are reared to adulthood by the brood parasite). This is somewhat surprising because it has long dead host queen's workers (host workers) and they then conduct been recognized for endoparasites and diseases that parasite-para- raids on neighbouring host from which they steal brood (larvae site interactions are likely to have profound effects on how parasitic and pupae) of the host ant species. The current generation of adult relationships evolve (Bull, 1994). For example, parasites may com- host raise the purloined brood to become the next generation pete with each other for the same host, which can affect aspects of of host workers (without a host queen, the parasitized colony cannot host-parasite interactions, such as the impact of the parasite on its produce its own host eggs or larvae). As a result, all parasite nests hosts (e.g., virulence, Read & Taylor, 2001). contain mixed-species colonies that comprise a parasitic dulotic Interactions between social parasites should be influenced by queen and her descendants, plus the stolen workers of numerous their population structure. For example, parasites might compete nearby host colonies, typically all from the same host species. among themselves for the same hosts, but such competition could Polyergus are widely reported to be hostile towards con- involve relatives in viscous populations. Interactions among kin specifics from neighbouring colonies (Bono, Gordon, Antolin, & could potentially lead to a reduction in the intensity of competition Herbers, 2006; Mori, Grasso, Visicchio, & Le Moli, 2001; Topoff, due to kin selection (Hamilton, 1964), although in some contexts, kin Lamon, Goodloe, & Goldstein, 1984; Trager, 2013), and the strongest competition might be expected to counterbalance any potential for evidence of this is the occurrence of intraspecific raids, where the kin selection (Gardner & West, 2004; Queller, 1994). raided colony is destroyed (in contrast to the more common raids on Conversely, population structure could reflect ecological in- pure host species colonies, where raided colonies typically survive) teractions among parasites. For example, in some species of social (Topoff et al., 1984). Similarly, Trager (2013) reports that when raids parasites, individual parasites specialize on one of several possible from neighbouring P. breviceps crossed paths, the interaction esca- host species, so ecological competition between parasites would be lated into a two-day battle and culminated in the destruction of one limited to parasites that share the same host species. Under some of the two parasite colonies. conditions, this type of host specialization can lead to the Other evidence suggests tolerance or at least less dramatic an- of genetically distinct host races, which are thought to be import- tagonism among conspecific Polyergus colonies. Some colonies do ant precursors to sympatric speciation (Buschinger, 1989; Gibbs manage to persist much closer to each other than expected based et al., 2000; Marchetti, 1998; Torres, Tonione, Ramírez, Sapp, & on their typical raid distance, despite the fact that the mechanics Tsutsui, 2018; Via, 2001). The term “host race” is probably overused of preraid host- searching by Polyergus scouts suggests such (Funk, 2012), given that the generally accepted definition requires neighbouring colonies must be aware of each other's proximity documenting several criteria (Drès & Mallet, 2002) that are often (Bono et al., 2006). Bono et al. (2006) suggest that directional and very difficult to confirm. However, even for species where not all temporal bias in raids from neighbouring nests may reflect a strat- criteria apply, and hence where sympatric speciation is unlikely, the egy of mutual avoidance among competing conspecific parasites. existence of host specialization and some population structure can However, distinct host preferences—and potentially concomitant have important consequences for parasite ecology and evolution. genetic differences—among neighbours may also influence the prox- Specifically, the existence of genetic differentiation among para- imity of colonies and their levels of intraspecific antagonism. In sum, site groups that use different hosts, despite some gene flow among Polyergus colonies exhibit widely varying levels of hostility towards them, is both consistent with host races and has ecological, social, conspecifics, and it is unknown how the interacting forces of host and evolutionary implications by itself. Genetically differentiated specialization and genetic structure influence these interactions. groups of interacting conspecific parasites using different hosts may While definitive examples of host races are lacking for socially have different mating constraints, dispersal patterns, and social in- parasitic ants, at least some of the criteria that define host races teractions than a panmictic population of the same parasites. have been documented for several dulotic species (Bono, Blatrix, Eusocial insects are an ideal group for examining the genetics Antolin, & Herbers, 2007; Goodloe & Sanwald, 1985; Goodloe, of social parasitism because they have complex societies and social Sanwald, & Topoff, 1987; Schumann & Buschinger, 1994, 1995), and parasitism is widespread (Bourke & Franks, 1995; Field, 1992; Heinze recent evidence suggests that they may occur in our study popula- & Keller, 2000; Hölldobler & Wilson, 1990). For example, ants in the tion as well (Torres et al., 2018). For all of these examples, a single genus Polyergus are obligate social parasites that can have many ant species parasitizes multiple species of host ants, but individual conspecific interactions, so their population structure may be par- colonies use only a single host species. This host fidelity has also ticularly relevant to understanding these interactions (Trager, 2013). been documented at our site (Torres et al., 2018) and is generally These ants show a form of social parasitism called dulosis (often common among Polyergus (Trager, 2013). “slave-making”, but see Herbers, 2007), which has evolved at least Host fidelity alone, while not sufficient to prove the existence of 10 times in ants (Beibl, Stuart, Heinze, & Foitzik, 2005; d'Ettorre host races, has important implications for dulotic ants. Parasites’ per- & Heinze, 2001). Dulosis is invariably characterized by two essen- ceptions of nestmates, kin, and conspecific competitors are all likely tial features of the parasite's life cycle: (a) a newly-fertilized dulotic to be profoundly influenced by the cuticular hydrocarbons of their 2052 | SAPP et al. host species of ant (Bos & d'Ettorre, 2012; Martin & Drijfhout, 2009; mtDNA markers. By using the same microsatellite markers on a much Topoff, 1990). Since parasites acquire both their chemical recogni- larger sample of colonies (82), we can test whether the reported dif- tion templates and their own cuticular hydrocarbons from host nest- ferences in number of genetic populations revealed by mtDNA and mates, parasites are likely to both choose mates from, and to show microsatellite markers is real (and perhaps a reflection of different competitive interactions with, other colonies that use the same host inheritance patterns of the two markers) or is an artifact of small species. Thus, host preference alone may influence both genetic sample size and the potentially highly related individuals (nestmates) population structure and the competitive interactions of neighbour- analysed. Specifically, evidence for three subpopulations would ing parasites independently. support the Torres et al. (2018) finding that maternal and paternal Relatedness is another factor that might affect parasite intraspe- patterns of gene flow differ for this parasite and would correspond cific interactions. Two behaviours that could lead to kin-structured to the three known host species accreta, F. argentea, and F. populations often occur in Polyergus species. First, mating behaviour subaenescens (Schär et al., 2018). Conversely, evidence supporting varies across the genus Polyergus, but observations at our site are four subpopulations would suggest that maternal and paternal gene consistent with the “female calling syndrome” (see pp. 144–146 of flow follow the same pattern, and that population structure is influ- Hölldobler & Wilson, 1990): new queens forgo flying altogether and enced by some unknown factor, possibly a cryptic 4th host species. instead attract males via pheromones. After mating monogamously on the ground near their natal nests, females remove their wings and search on foot for host colonies to infiltrate. Consequently, mat- 2 | MATERIALS AND METHODS ing and parasitic nest founding happen close to a new queen's natal nest. This should lead to relatedness among neighbouring colonies 2.1 | Study site and subjects and possibly a pattern of average relatedness that decreases with the distance between nests. Second, conspecific colony destruction We studied a population of Polyergus mexicanus on the east slope affects the spacing of parasitic nests and, if kinship influences the of the northern Sierra Nevada Mountains, approximately 20 miles pattern of colony destruction, neighbouring colonies are likely to be north of Lake Tahoe at the Sagehen Creek Field Station (“SCFS", a closely related to each other. University of California Natural Reserve; 39.432181, −120.241263). In this study, we investigate patterns of population structure Note that P. mexicanus at this site was previously identified as P. with respect to host use and relatedness in the dulotic ant Polyergus umbratus, characterized by a long and often convex mesonotum mexicanus. First, we test the idea that host preference influences compared to P. mexicanus. Although P. umbratus was recently syn- genetic structure, spatial ecology, and intraspecific interactions. We onymized with P. mexicanus (Trager, 2013), recent genetic work indi- quantify the population structure and relatedness of neighbouring cates that P. umbratus is actually a distinct species, so the name may colonies in a spatially contiguous range of P. mexicanus nests and soon be resurrected (J. C. Trager, personal communication). contrast this population structure with host species use, nest loca- The site comprises a variety of habitats, but nests were typically tion, and intraspecific interactions. Second, we quantify the poten- found within 200 m of dirt roads in disturbed (mechanically thinned tial for kin population structure in two ways: (a) we determine the for fire control) mixed-conifer forest on the south-facing slope of relationship between the relatedness of pairs of colonies and their the Sagehen Creek drainage basin. Nests were often found associ- distance from each other; and (b) we determine the frequency with ated with downed tree trunks, stumps of harvested trees, or the root which colonies had any first order relatives as neighbours. structure of common understory plants such as Ceanothus prostratus Our initial goal was to focus on kin structure in our population, and Wyethia mollis. but it became clear that the existence of host races could confound The elevation of study populations ranged from 1,930 to 2,125 m such an investigation. Specifically, ignoring genetic structure from over a contiguous area of approximately 9 km2. We estimate a den- host races could inflate apparent kin structure because colonies that sity of P. mexicanus nests at 8.4 per 100 m2, which is greater than share the same host species could have increased genetic similarity any we are aware of elsewhere in the literature for any species of (and hence apparent relatedness) relative to the entire population, Polyergus. While P. mexicanus colonies at SCFS are only known to due to genetic isolation or selection from the history of host use and parasitize F. accreta, F. argentea, and F. subaenescens, there are ap- not from true relatedness. We therefore expanded the scope of the proximately 20 species of Formica at our site, many of which are study to include an analysis of host races but note that we are not quite similar in habitat and appearance to the three most common the first to examine host races in this species. Torres et al. (2018) host species. previously examined the existence of host races occur in this same population, but with a much more restricted sample of colonies (18 colonies with mtDNA, 10 colonies with microsatellite DNA) than 2.2 | Field sampling methods and design we used (82 colonies, seven of which were shared by both studies). Torres et al. (2018) found evidence of three distinct genetic subpop- From 2008 to 2010, we searched for raids and nests on two 1-hectare ulations using the same biparentally inherited microsatellite mark- study plots. To better characterize the genetic diversity of the popula- ers we use here but found evidence of four subpopulations using tion of P. mexicanus, we also searched a ~15 km2 area for nests along SAPP et al. | 2053 roads and trails throughout the reserve in 2010. In 2011, we conducted MiniPrep Zymo kits according to the manufacturer's instructions. daily observations on four additional smaller (2,500 m2) focal plots to We amplified the DNA with PCR using six primers developed provide more independent observations of unique pairs of interact- by Bono et al. (2007): Pol1, Pol2, Pol3, Pol4, Pol5, and Pol12. We ing colonies for behavioural studies and broaden our genetic sample modified these original primers to use an M-13 dye-tagging proto- of parasite colonies for this study. All six plots were centered on a P. col (Schuelke, 2000). Each PCR was labelled with one of Applied mexicanus nest and were chosen because of the high density of sur- Biosystems DS-33 Dyes (LIZ, 6-FAM, VIC, NED, PET). rounding P. mexicanus nests in the area as revealed through preliminary The amplification process for all six loci differed only in anneal- pilot searches for nests. Our observations and collection of specimens ing temperature. For all loci, extracted DNA was initially denatured were not limited to plot boundaries: when we detected raids and nests at 95˚C for 5 min, then run through 36 cycles, each of which con- near but outside plot boundaries, we included them in this study. sisted of additional denaturing at 95˚C for 30 s, 30 s at one of two We searched for raids at these plots daily for at least a month annealing temperatures, and 30 s of extension. After these 36 cy- during the peak of the raiding season (typically during July) and used cles, there was a final extension step of five minutes at 72˚C be- the conspicuous raids to locate the inconspicuous nests of both the fore samples were stored at 4˚C. The annealing temperature was mixed-species P. mexicanus colonies and the colonies of their host 58˚C for Pol1, Pol4, and Pol12, and 53˚C for Pol 2, Pol 3, and Pol 5. Formica species. We did not attempt to estimate the relative abundance Amplified DNA was preserved in HiDi Formamide and sent to the of nests of the three known Formica species in this study because our University of California at Berkeley Sequencing Facility for microsat- study is focused on interactions among parasites and because Formica ellite fragment size analysis using Applied Biosystems 3730XL DNA nests are cryptic unless being raided. We collected data on the scale, Analysers and LIZ size standard. We determined peaks and bins of frequency, and nature of intraspecific parasite interactions by measur- each locus on the resulting electropherograms using geneious version ing the distance of raids, recording the relative locations of all parasite 6.0 (Kearse et al., 2012). nests, and making special note of any raids that crossed any other ac- tive raids and intraspecific raids (i.e., raids from one parasite nest to another). We collected between 1–12 individuals from 82 nests for a 2.4 | Population structure analysis total of 397 P. mexicanus female workers and six males for genomic

DNA extraction. Live ants were frozen and preserved in 95% ethanol. To estimate population assignment of individuals, we used structure To assess host species identity, we collected host Formica workers (Pritchard, Stephens, & Donnelly, 2000) which uses Bayesian tech- either from a focal P. mexicanus colony or from a Formica nest raided niques to form clusters of individuals that best meet the assump- by that focal P. mexicanus colony. Because it is well established that tions of Hardy-Weinberg equilibrium and linkage disequilibrium. individual P. mexicanus colonies use a single host species, we often We tested hypotheses for K = 2–10 populations with 50 independ- opted to collect Formica specimens and raiding Polyergus at the site of a ent runs for each hypothesized K value (400 total runs). We used raided nest to minimize disturbance to our long-term focal objects: the a burn-in length of 10,000 followed by 100,000 Markov Chain parasitic Polyergus colonies. For each parasite colony, we attempted Monte Carlo (MCMC) repetitions. We set λ = 1 to assume that al- to collect at least three host Formica workers, which we mounted ac- lele frequency could vary independently among subpopulations cording to museum standards for species identification. We used the (a standard assumption when the K value is unknown) and did not dichotomous keys developed by Francœur (1973) as well as several use any prior population information (POPFLAG = 1). We used the characters known to be locally diagnostic for the different host species online tool structure harvester (Earl & vonHoldt, 2012) to collate (P. S. Ward and C. W. Torres, personal communication) to determine these results to determine which hypothesized K was most likely, the species identity of the Formica host workers. However, characters using the Evanno, Regnaut, and Goudet (2005) method. We then were sometimes ambiguous and Formica species within the subae- used clumpp (Jakobsson & Rosenberg, 2007) to create a consensus nescens-group (as are the three host species at SCFS) are notoriously data set that used the data from all the structure simulations of the difficult to identify (Glasier, Acorn, Nielsen, & Proctor, 2013; Mackay best supported K values. Because our inclusion of nestmates poten-

& Mackay, 2002). To avoid making classification errors we assigned tially violates structure’s assumption that all samples are unrelated ambiguous individuals as “cf. Formica subaenescens” and “cf. Formica individuals, we repeated the above steps with only one individual argentea”. For all analyses, we checked to see if inclusion or exclusion from each colony to confirm that the best supported K, and the sub- of these ambiguous individuals changed the results. For analyses un- population assignments of individuals did not change. We analysed affected by inclusion of these individuals, we report the statistics that the goodness of fit between our estimated genetic subpopulations include these individuals assigned to their most likely species. and host species groupings we described based on the morphology of host species for each P. mexicanus colony. Finally, we used the

graphical software distruct (Rosenberg, 2004) to visually contrast 2.3 | Microsatellite protocol these genetic populations with the three host species groupings we determined by examining host Formica workers. We extracted genomic DNA from the collected P. mexicanus work- Since males are haploid, we coded their genotypes as miss- ers and six males using either Qiagen DNEasy Kits or Quick-gDNA ing a second allele at each locus for the structure analysis, as 2054 | SAPP et al. recommended (Pritchard et al., 2000). After the structure analysis, Because kinship estimation relies on deviations from the assumption we removed 11 individuals (including all six males) that either had of panmixia, we restricted these calculations to all individual work- less than 50% of loci amplify or had otherwise incomplete or in- ers who were assigned to the same genetic subpopulations, as esti- consistent data from subsequent analyses on relatedness. We also mated from the structure analysis. Within each subpopulation, we excluded nine other individuals that had a maximum subpopula- calculated the relatedness between all possible pairs of sampled P. tion assignment of less than 80% from relatedness analyses, be- mexicanus workers. We determined the relatedness between each cause relatedness calculations are only sensible when computed pair of colonies by averaging all possible pairwise relatedness values between members of the same subpopulation; however, including of each nestmate from the first colony with each nestmate from the all individuals did not qualitatively change our results here. We second colony. also excluded these nine individuals from tests of the relation- We calculated the Pearson product-moment correlation for the ship between host species and subpopulation assignment, be- pairwise relatedness of colonies and the distance between nests for cause these analyses required unambiguous assignment into one all pairs of colonies, which included nest pairs as far as 4 km away subpopulation. from each other, a distance we considered reasonable for the dis- We checked for null alleles, allelic dropout, and stutter in each persal of winged males. To focus on the effects of female dispersal of subpopulations we detected via the structure analysis using mi- (which is on foot) and intercolony interactions such as conspecific crochecker (Van Oosterhout, Hutchinson, Wills, & Shipley, 2004) raids, we repeated this analysis with only colony pairs with nests software that compares molecular data to Hardy Weinberg ex- that were less than twice the maximum observed distance of host pectations and thus requires an estimate of population boundaries raids, 155 m. among samples. Because microchecker assumes that samples within To assess the potential for kinship to influence interactions a population are from unrelated individuals, we randomly chose only among colonies with neighbouring nests, we calculated the propor- one individual per sampled colony for the microchecker data set. tion of colonies that had at least one highly related neighbour within However, this random subsample excluded seven rare alleles that the maximum observed raiding distance (77.5 m) of a focal nest. We were present in the complete sample, so we also added six individu- considered a relatedness value >0.375 as highly related; this is the als from colonies that were already represented to ensure that these expected relatedness between workers of one colony and workers alleles were included in our analyses. In other words, we allowed a of a new queen arising from that colony. We present these results slight violation of the software's assumption regarding relatedness both as a proportion of all 82 colonies observed, and a proportion of of samples in favor of including all alleles. We found evidence of null the 44 colonies that have at least one neighbour. alleles for only one locus (Pol 1) in one of our estimated subpopu- lations (Population 3), and we used the “Brookfield 2” corrected al- lele frequency provided by microchecker for subsequent relatedness 3 | RESULTS analysis (Brookfield, 1996). 3.1 | Population structure

2.5 | Comparing population structure to host We found the strongest evidence for three subpopulations (Delta species use K = 849 for K = 3, Table 1). The hypothesis of four subpopulations was poorly supported by the genetic structure analysis (Table 1) and To assess the relationship between host Formica species use and ge- did not align well with patterns of host species use (Figure 1c). The netic subpopulation, we first calculated a colony-level subpopulation three well-supported subpopulations were very distinct, with most probability of assignment score for each colony by taking the high- individuals very clearly assigned to one of the three populations est subpopulation probability score assigned by structure to each (Figure 1a). Evidence for gene flow between subpopulations was nestmate and averaging them together. In all cases, nestmates had minimal: we detected only 12 individuals out of 397 sampled whose the highest probability of assignment to the same subpopulation, in- membership coefficients were less than 0.8 for all three clusters, and dicating that a colony average is biologically meaningful. We tested three of these were individuals with poor microsatellite data (see an- the dependence of host species use on a parasite's genetic subpopu- notations in Figure 1a). lation using a chi-squared contingency test. The number of genetic subpopulations estimated by our pop- ulation structure analysis (K = 3) equaled the number of known host Formica species parasitized by P. mexicanus at SCFS. Genetic 2.6 | Relatedness analysis subpopulations of P. mexicanus colonies were correlated with the host Formica species they used (Contingency test, χ2 = 57.1, df = 4,

We used kingroup (Konovalov, Manning, & Henshaw, 2004) to meas- p < .001), but this relationship was not perfect. Subpopulation 3 in ure pairwise relatedness between all possible pairs of individuals particular showed little evidence of host specialization compared within each subpopulation detected in our genetic structure analysis. to the other two subpopulations. While colonies in Subpopulations SAPP et al. | 2055

1 and 2 predominantly used F. argentea and F. subaenescens hosts, 3.2 | Relatedness between parasite colonies and respectively, Subpopulation 3 used all three hosts with moderate among individuals within colonies frequency (Table 2). The global average relatedness of all non-nestmates was close to zero (R ± SD = 0.03 ± 0.27; n = 26,410 unique pairs of non-nestmate TABLE 1 Evanno method statistics for each hypothesized number of subpopulations (K) workers), indicating that most colonies are not closely related to each other. However, 75 colonies pairs (out of 1,363 possible combina- Mean SD of Delta tions of colonies) had an average relatedness equal to or greater than K Reps LnP(K) LnP(K) Lnʹ(K) |Lnʺ(K)| K 0.375; the expected relatedness between workers of one colony and 2 50 –7317 81.51 – – – workers of a new queen arising from that colony (Figure 2a). Eleven 3 50 –6550 0.67 767 568 849.31 of the 130 (8.5%) unique pairs of colonies with nests within 155 m 4 50 –6351 78.96 198 67 0.86 of each other had a relatedness value of 0.375 or higher (Figure 2b). 5 50 –6221 125.11 130 16 0.13 We found that 44 colonies of the 82 colonies in our genetic sam- 6 50 –6107 93.60 114 5 0.06 ple had at least one neighbour from their subpopulation within the 7 50 –5988 43.82 119 35 0.79 maximum raiding distance (77.5 m). Of these 44 colonies, 16 (35.4%) 8 50 –5903 90.30 85 23 0.25 had a neighbouring colony whose average relatedness was 0.375 or 9 50 –5842 97.08 62 7 0.08 greater. There was a weak but significant negative correlation between 10 50 –5773 99.37 69 – – relatedness of colonies and the distance of their nests from each Note: LnP(K) is the log-probability of K. SD is standard deviation. Lʹ(K) is other over local distances (up to 155 m; Table 3). However, there was the rate of change of the likelihood distribution. |Lʺ(K)| is the absolute value of the second order rate of change of the likelihood distribution. no such correlation in an analysis that included pairwise comparisons Delta K is calculated as |Lʺ(K)|/SD of LnP(K). at all distances (Table 3). Separate analyses of each subpopulation

FIGURE 1 Structure plots for Polyergus mexicanusstudy population. Plots compare best supported number of genetic subpopulations, K = 3 (a, b) to K = 4 (c). Shading of main bars represents subpopulation (for K = 3: black is Subpopulation 1, dark grey is Subpopulation 2, light grey is Subpopulation 3. K = 4 populations were not named). The alternating black and white bar along the bottom indicates colony membership: a contiguous shading indicates that individuals belong to the same colony. (a) Individuals and colonies sorted by subpopulation assignment. The 18 individuals that were removed from subsequent analyses are noted with three symbols: along the top, solid triangles (▼) identify 12 individuals whose highest membership coefficient was <0.8, and hollow triangles (▽) indicate three individuals that had <50% scorable microsatellite loci. Along the bottom, six males are indicated by diamond (◆) symbols. (b) Individuals and colonies sorted by host species use. Gradient shaded bars along bottom group colonies that use the indicated Formica host species. (c) Individuals and colonies arranged as in (b), but with K = 4 2056 | SAPP et al. revealed that only Subpopulation 1 had a significantly negative cor- involving colonies of the same subpopulation, relatedness values relation between distance and relatedness, and that this correlation ranged from –0.23 to 0.39. existed even when all distances were considered (Table 3). Six pairs of raids crossed each other from five unique pairs of Within colonies, relatedness was high: the global average pair- colonies. There were no signs of aggression or antagonism for any wise relatedness of nestmates was close to the 0.75 value expected of these crossing raids, in stark contrast to all intraspecific raids we for full sibling haplodiploid sisters (R ± SD = 0.71 ± 0.24, n = 939 observed. We had complete genetic data and host species data for unique pairs of nestmates). one pair, and host species data for three pairs. One pair was from the same subpopulation (Subpopulation 1) and had a colony-level re- latedness value of 0.181. The three other pairs were from different 3.3 | Parasite intraspecific interactions subpopulations and all but one of these used different host species. The raids that crossed but used the same host species (F. subae- We observed 778 raids from the parasite colonies involved in this nescens) were from Subpopulations 2 and 3. study. We combined these observations with those from other ob- servations at this site (i.e., colonies with no genetic samples) for a total of 862 observed raids with the maximum raid distance of 77.5 4 | DISCUSSION (as used in above analyses), and an average raid distance of 19.2 m (SD = 13.6). Nests of parasitic colonies were frequently closer to The relationship between genetic structure, parasite host use, popu- each other than these raiding distances, and there was no apparent lation viscosity, and intraspecific interactions among these parasites clustering by subpopulation or host use (Figure 3). is more complex than predicted. We found strong evidence that host Among these raids, 11 were intraspecific raids that occurred species use restricts gene flow, and that genetic subpopulations are between seven unique pairs of colonies (i.e., some intraspecific associated with the use of specific host species. However, the ge- raids happened repeatedly between the same pair of colonies). We netic subpopulations we detected do not perfectly match parasite had both genetic data and host species data for both colonies in host use. At a local level, the relatedness patterns we observe be- four of the seven unique pairs of colonies. In all cases, the inter- tween pairs of colonies do not indicate high viscosity: neighbour- acting colonies shared host species. In one of these four cases, the ing colonies are often not closely related, nor necessarily from the raided and raiding colonies belonged to different subpopulations same subpopulation, nor using the same species of host (Figure 3). (Subpopulation 1 raided Subpopulation 3). For the three other cases However, while most neighbouring colonies of the same subpopula- tion are not close relatives, colonies that are both closely related and spatially close enough to interact with each other occur regularly, TABLE 2 The relationship between host species use and and this could affect the nature of interactions between some colo- parasite genetic subpopulation nies. The relationship between distance and relatedness between Genetic subpopulation pairs of colonies within raiding distance of each other was weaker than expected from our observations of the mating behaviour of P. Host species 1 2 3 Totals mexicanus queens and intraspecific raiding behaviour of P. mexicanus Formica accreta – – 17 17 workers at SCFS. The interactions we observed between pairs of 23 2 3 28 parasite colonies—raids on each other and raids from two different Formica subaenescens 1 8 13 22 colonies that crossed—were extremely infrequent in comparison to Totals 24 10 33 67 normal raiding behaviour. While the rareness of these observations Note: Frequencies of observed Polyergus colonies that occur in each prohibits formal analysis of their relationship to other factors, it genetic subpopulation contrasted with frequencies that use each host seems clear that conspecific raids only occur between colonies that Formica species. use the same host species, and that no clear pattern exists between

FIGURE 2 Histograms of average relatedness of pairs of colonies for highly related colonies at two spatial scales. Only colony pairs with average R > 0.375 are included. (a) All possible pairs of colonies at our study site, irrespective of distance between their nests. (b) Pairs of colonies with nests <155 m from each other SAPP et al. | 2057

TABLE 3 Correlations between Genetic Sample relatedness and distance for different Distances subpop size df p t Correlation subpopulations at two spatial scales All All 1,782 1,780 .06 –1.88 –0.04 <155 m All 134 132 .04 –2.08 –0.18 All 1 627 625 <.01 –5.02 –0.20 All 2 104 102 .12 1.59 0.16 All 3 1,051 1,049 .38 0.87 0.03 <155 m 1 48 46 .02 –2.47 –0.34 <155 m 2 12 10 .22 –1.30 –0.38 <155 m 3 74 72 .76 –0.30 –0.04

Note: Subpop, subpopulation. Statistics shown are Pearson product moment correlations for relatedness versus distance for all unique pairs of colonies.

FIGURE 3 Distribution of Polyergus mexicanus nests in the study area. Map showing the locations of nests located in and around the borders of the two most intensively studied one ha plots established in 2008, which are represented by dashed lines and shown in detail in the two lower panels. Shape indicates host species: F. accreta = circles (n = 8), F. argentea = squares (n = 22), F. subaenescens = triangles (n = 25), and unidentified host species = asterisks (n = 27). The shading of each marker indicates if the colony was assigned to Subpopulation 1 (black; n = 30), Subpopulation 2 (white; n = 13), or Subpopulation 3 (grey; n = 39)

either type of intraspecific raid interaction and genetic subpopula- to play a role in sympatric speciation of the parasites. These criteria tion or relatedness. include (a) sympatric subsets of a parasitic species that use different When a parasitic species uses several different host species, co- host species; (b) statistically significant genetic differences among the evolution between host and parasite can, under some conditions, lead parasites using different hosts; (c) the genetic differences must extend to the evolution of distinct genetic “host races” (Gibbs et al., 2000; beyond loci that affect host preference; and (d) the genetic differences McCoy, Boulinier, Tirard, & Michalakis, 2001). Jaenike (1981) outlined must not be entirely the result of natural selection on the current gen- the criteria for distinguishing between simple differences in current eration. More recently, Drès and Mallet (2002) expanded these criteria host preferences and true genetic host races that have the potential to include the following: (a) greater genetic differences among host 2058 | SAPP et al. races in sympatry than between individuals using the same host in al- hybridization among the closely related Formica host species. While lopatry; (b) parasite mate choice that is correlated with host choice; and hybridization could explain the existence of the ambiguous host (c) gene flow between races in sympatry. species we identified as F. cf. argentea and F. cf. subaenescens, and While these criteria are necessary to fully demonstrate that an the use of putative F. subaenescens hosts by two distinct Polyergus observed host-parasite association represents a step on the path subpopulations, we found exceedingly little evidence of gene flow to sympatric speciation, the difficulty in documenting all six criteria between parasite subpopulations in the patterns of their genetic has made true examples of host races rare (Fitzpatrick, Fordyce, & structure, in contrast to what we would expect if hosts were hybrid- Gavrilets, 2008; Via, 2001). Evidence for host races does exist for izing. If cryptic host species occurred commonly, we would expect some parasite groups (e.g., ectoparasites [McCoy et al., 2001]) but to find more subpopulations than known host species, but in fact we similarly complete evidence in social parasites has rarely been found strong support for three parasite subpopulations, the same as studied (but see Fanelli et al., 2005), and not previously reported. the number of known host species. Though not sufficient to prove host race formation, cases of clear Assuming our determination of host species identity is accu- genetic differentiation among parasites using different hosts have rate, the observed relationship between genetic subpopulation implications for the ecology and evolution of host-parasite interac- and host Formica species suggests a more complex pattern of host tions, even when there are mechanisms in place that prevent spe- use than the one-to-one host-species-to-parasite subpopulation ciation. One example is the avian brood parasite, Cuculus canorus, pattern predicted by the host-race hypothesis. Genetic subpop- in which speciation is prevented by cross-host mating by males, and ulations 1 and 2 provide evidence reasonably consistent with the genes responsible for host-specialization reside on the female sex existence of P. mexicanus host races specializing on F. argentea and chromosome (Gibbs et al., 2000; Marchetti, 1998). F. subaenescens, respectively. Despite this apparent host specializa- In this context, the recent other work at SCFS showing a clear tion by Subpopulation 2 on F. subaenescens, a parasite colony using relationship between parasite genetic structure and host species F. subaenescens was more likely to belong to Subpopulation 3 than use is noteworthy, especially since it also documents greater ge- Subpopulation 2. Subpopulation 3's relatively even use of all three netic similarity between allopatric populations using the same host host species is evidence against potential host race formation in that than between sympatric populations using different hosts (Torres subpopulation (Figure 1b and Table 2). et al., 2018). Our results corroborate parts of this finding, while One interpretation of this pattern is that that Subpopulation adding some complications. First, we confirm that there are three 3 may be more generalist than the other two subpopulations and clearly defined genetic subpopulations at SCFS. This clear pattern is thus able to use F. accreta, F. subaenescens, and to a lesser extent, supported with 397 ants from 82 distinct colonies and suggests that F. argentea as hosts. Host specificity may be caused by limited par- the difference Torres et al. (2018) found between biparental mark- asite dispersal or host-specific adaptation (Timms & Read, 1999). ers (three subpopulations) and maternally inherited markers (four Given the general proximity of nests of parasite colonies using the distinct clades) represents a real difference in inheritance patterns, different three host species at our site, and the intimate levels of which suggests differing maternal and paternal gene flow, or differ- deceptive chemical signaling by the parasites that must occur for ing rates of evolution between the two markers. Indeed, our results dulosis to be successful (d’Ettorre & Heinze, 2001), host-specific ad- indicate that the hypothesis of four genetic subpopulations is a poor aptation seems a more likely driver of host specificity. Because host fit with both our biparental microsatellite data (Table 1) and our host specificity has coevolutionary consequences for host resistance use data (Figure 1c). However, our results do not indicate that ge- and parasite virulence (Little, Watt, & Ebert, 2006), determining netic structure maps onto host use as cleanly as suggested by Torres the extent of host specificity in different parasite subpopulations, et al. (2018). This could be because the broader sample of parasite as well as its causes and consequences, is a valuable direction for colonies here better captures true variable relationships between future work. host use and genetics in this parasite population, suggesting that host Despite the diversity of hosts used by Subpopulation 3, one species use is only one factor influencing genetic structure among pattern does suggest its genetic structure is connected to host spe- other unknown factors. Alternatively, our results could indicate that cialization: it is the only subpopulation we observed using F. accreta the morphological indicators we used to determine host species are hosts (Table 2). This suggests that Subpopulation 3 may have co- unreliable. Indeed, these Formica are infamously difficult to correctly evolved to use F. accreta hosts historically, but still has the capac- identify (Glasier et al., 2013; Mackay & Mackay, 2002), and cryptic ity to use the other two hosts when F. accreta hosts are limiting. species and hybrids are common in the genus (Beresford et al., 2017; This hypothesis is supported by the fact that F. accreta is the rarest Bernasconi, Cherix, Seifert, & Pamilo, 2010; Seifert, 2009; Seifert, host we observed, yet Subpopulation 3 is the most abundant para- Kulmuni, & Pamilo, 2010). A modern phylogenetic analysis of host site population: Subpopulation 3 parasites may prefer F. accreta, but Formica species could help resolve this issue and help identify any can make the “best of a bad job” with F. argentea or F. subaenescens currently cryptic species. colonies when they are abundant while F. accreta colonies are rare. The genetic evidence, however, does not suggest that the in- Future work that better documents the abundances and locations of consistent match between genetic subpopulations and host species free-living host Formica nests and that measures fitness proxies of use is caused by the existence of cryptic Formica host species, nor parasite colonies as a function of the interaction between genetic SAPP et al. | 2059 subpopulation and host species will be an important next step in intraspecific raids, relatedness values between such pairs of colo- testing this idea. nies varied from colonies that were less related than the population The mating behaviour of eusocial organisms has strong effects average to colonies that appeared to be kin. This variation in related- on their dispersal, reproductive output, social evolution, and ecology ness values suggests that kinship is not an important determinant of (Liautard & Keller, 2001; Vitikainen, Haag-Liautard, & Sundström, parasite raids on other parasite colonies and that local competition 2015). At SCFS, the mating behaviour of new P. mexicanus queens is sufficient to overcome kin-selected altruism for these parasites, as is characterized by short dispersal distances because queens do not many models of kin selection under limited dispersal indicate (Griffin fly, but attract flying males (“female calling syndrome”; Hölldobler & & West, 2002; West, Pen, & Griffin, 2002). Wilson, 1990), and then mate monogamously with a single male. We In contrast to what has been reported for other sites expected that such limited dispersal by queens would lead to pop- (Trager, 2013), none of the six pairs of raids we observed crossing ulation viscosity, where individuals in spatial proximity would tend each other exhibited any discernible hostile interactions. Although to be closely related. In addition to mating patterns, raids among we do not have sufficient relatedness or subpopulation data to cor- parasites on each other's nests could also structure populations if relate with these passive crossing raids, they are striking events that the raids were biased towards unrelated individuals: local removal strongly suggest some form of intraspecific tolerance. It is likely that of unrelated individuals should further contribute to relatedness of crossing raids are capable of hostility but choose peace, because two neighbouring parasite colonies. While we did find that relatedness lines of evidence suggest Polyergus raids do indeed have the capacity between pairs of colonies did decrease with distance over the prob- to opportunistically attack each other when their raids cross in other able range of P. mexicanus queen dispersal, the correlation was weak. contexts. First, Trager (2013) reports similar crossing raids quickly Strangely, this pattern seems driven by entirely by Subpopulation devolving into intraspecific hostility on par with intraspecific raids. 1, and the relationship persisted when all distances were consid- Unfortunately, we do not know the host-species or relatedness con- ered for this subpopulation (Table 2). We have no a priori reason to text in which this conspecific interaction occurred, but it would be suspect that one subpopulation would have different dispersal or informative to know how many host species were common at that mating behaviour than the others, even if we assume that subpopu- site. Second, as Topoff et al. (1984) has reported and we have ob- lations represent host races. Subpopulation 1 did have had a tighter served directly, Polyergus raiders appear to be flexible in the targets correlation with one host species (F. argentea) than the other two they choose during raids, often encountering other targets on the subpopulations. Future work on host quality and the ecological and way to the original Formica nest their scout was leading them to. behavioural differences among parasites that use different hosts This means that raiders are capable of “switching” from transit be- may illuminate why these subpopulations have differing relatedness haviour to actual raiding (i.e., attacking a nest) at any moment, and patterns across the landscape. they are not behaviourally constrained to keep running until they Many ants have viscous populations, due to limited dispersal reach a target nest. Both these examples suggest that Polyergus are (Chapuisat, Goudet, & Keller, 1997; Pamilo, Gertsch, Thorén, & behaviourally capable of impromptu attacks on rivals they meet Seppä, 1997; Sundström, Seppä, & Pamilo, 2005). While viscous while raiding, so our finding that they can also abstain from attack- populations were once thought to promote kin-selected altruism, it ing conspecifics is interesting. One tantalizing possibility for further is now appreciated that this altruism may be counteracted by the investigation is that host similarity, possibly including relatedness, increased kin competition that low dispersal engenders (Griffin & affects parasite nestmate recognition such that raids that cross rec- West, 2002; Queller, 1994). While we find weak evidence support- ognize each other as nestmates. ing population viscosity overall, we found that enough colonies have Parasites’ interactions with each other and their hosts have a closely related neighbour to make kinship potentially relevant to implications for speciation, co-evolution, virulence, population some of the intraspecific interactions between neighbouring para- growth, and competition (Bull, 1994; Buschinger, 1989; d’Et- site colonies. torre & Heinze, 2001; Foitzik, DeHeer, Hunjan, & Herbers, 2001; Some of these parasitic ants’ most striking intraspecific interac- Thompson, 2005). We show that patterns of host use affect gene tions—raids on each other's nests and raids from adjacent nests that flow in a sympatric parasite population, but that not all of this ge- cross without aggression—are also among the most infrequent. We netic structure can be explained by host use. We also demonstrate have too few observations on these interactions to permit statistical the opportunity for kin-selected interactions but find little evidence analysis, but the initial patterns are not consistent with predictions that they are more important than local competition in this system. we might make based on kin selection or subpopulation structure, Our results are a valuable starting point for future work on para- though host species seems relevant. All intraspecific raids oc- site-parasite interactions, parasite and host genetic structure, para- curred between colonies using the same host species, as predicted site mating and dispersal behaviour, parasite competition, and host by the known host fidelity exhibited by Polyergus colonies at this phylogenetics. site. Specifically, these intraspecific raids may be adaptive because raiders obtain additional brood of their desired host species, and ACKNOWLEDGEMENTS because they remove or harm a neighbouring competitor for that We thank the following field and laboratory assistants who as- brood. In contrast to the host species status of colonies involved in sisted in finding nests, collecting specimens, and assisting with 2060 | SAPP et al. molecular work: Skyler Hackley, Matt Freitas, Yasmine Akky, Formica sp. cf. argentea): Host specificity and coevolutionary dynam- ics. Biological Journal of the Linnean Society, 91(4), 565–572. https:// Andrew Frostholm, Beatriz Nobua Behrmann, Mallory Hee, doi.org/10.1111/j.1095-8312.2007.00817.x Jessica Simms, Steven Cooper, Hannah Nolan, and Don Brown. Bono, J. M., Gordon, E. R., Antolin, M. F., & Herbers, J. M. (2006). We thank Phil Ward, Candice Torres, and James Trager for sharing Raiding activity of an obligate (Polyergus breviceps) and two faculta- their taxonomic expertise and natural history insights that were tive (Formica puberula and F. gynocrates) slave-making ants. Journal instrumental in the design of this study. We thank UC Berkeley's of Insect Behavior, 19(4), 429–446. https://doi.org/10.1007/s1090​ 5-006-9035-2 Sagehen Creek Field Station, and the University of California Bos, N., & d'Ettorre, P. (2012). Recognition of social identity in Natural Reserve System for providing the ideal natural environ- ants. Frontiers in Psychology, 3, 83. https://doi.org/10.3389/ ment to conduct this study. We also thank station managers Jeff fpsyg.2012.00083 Brown and Faerthen Felix for facilitating our research and resi- Bourke, A. F. G., & Franks, N. R. (1995). Social evolution in ants. Princeton, NJ: Princeton University Press. dency there for the duration of this study. We thank Neil Tsutsui, Brooker, M. G., & Brooker, L. C. (1989). The comparative breeding be- Giacomo Bernardi, and Barry Sinervo for feedback on experimen- haviour of two sympatric cuckoos, Horsfield’s Bronze-Cuckoo tal design and earlier versions of this manuscript. This work was Chrysococcyx basalis and the Shining Bronze-Cuckoo C. lucidus, in supported by UC Natural Reserve System Mathias grants, a UC Western Australia: A new model for the evolution of egg morphology and host specificity. Ibis, 131, 528–547. Santa Cruz STARS grant, UCSC Ecology and Evolutionary Biology Brookfield, J. F. (1996). A simple new method for estimating null allele department graduate summer research grants, and a Sigma Xi frequency from heterozygote deficiency. Molecular Ecology, 5(3), grant-in-aid of research. 453–455. https://doi.org/10.1046/j.1365-294X.1996.00098.x Bull, J. J. (1994). Virulence. Evolution, 48(5), 1423–1437. https://doi. org/10.1111/j.1558-5646.1994.tb02185.x​ AUTHOR CONTRIBUTIONS Buschinger, A. (1989). Evolution, speciation, and inbreeding in the par- J.R.S. designed and performed the research, analysed the data, and asitic ant genus Epimyrma (, Formicidae). Journal wrote the paper. J.Y. performed the research and edited the paper. of Evolutionary Biology, 2(4), 265–283. https://doi.org/10.104 B.L. designed the research and wrote the paper. 6/j.1420-9101.1989.2040265.x​ Chapuisat, M., Goudet, J., & Keller, L. (1997). Microsatellites reveal high population viscosity and limited dispersal in the ant Formica paralu- DATA AVAILABILITY STATEMENT gubris. Evolution, 51(2), 475–482. All data including microsatellite genotypes, estimated relatedness Davies, N. B. (2010). Cuckoos, cowbirds and other cheats. London, UK: T. values, nest locations, host species determinations, calculated dis- & A.D. Poyser. tances between parasite nests, and raid distances are openly avail- d'Ettorre, P., & Heinze, J. (2001). Sociobiology of slave-making ants. Acta Ethologica, 3(2), 67–82. https://doi.org/10.1007/s1021​10100038 able in Dryad at: https://datad​ryad.org/stash​/datas​et/doi:10.5061/ Drès, M., & Mallet, J. (2002). Host races in plant-feeding insects and their dryad.wstqj​q2h9 (Sapp, Yost, & Lyon, 2020). importance in sympatric speciation. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 357(1420), 471– ORCID 492. https://doi.org/10.1098/rstb.2002.1059 Earl, D. A., & vonHoldt, B. M. (2012). STRUCTURE HARVESTER: A Joseph R. Sapp https://orcid.org/0000-0002-2231-9878 website and program for visualizing STRUCTURE output and imple- Jenn Yost https://orcid.org/0000-0002-3340-2757 menting the Evanno method. Conservation Genetics Resources, 4(2), Bruce E. Lyon https://orcid.org/0000-0001-8733-9944 359–361. https://doi.org/10.1007/s12686-011-9548-7​ Evanno, G., Regnaut, S., & Goudet, J. (2005). Detecting the number of clusters of individuals using the software STRUCTURE: A sim- REFERENCES ulation study. Molecular Ecology, 14(8), 2611–2620. https://doi. Aron, S., Passera, L., & Keller, L. (1999). Evolution of social parasit- org/10.1111/j.1365-294X.2005.02553.x ism in ants: Size of sexuals, sex ratio and mechanisms of caste Fanelli, D., Henshaw, M., Cervo, R., Turillazzi, S., Queller, D. determination. Proceedings of the Royal Society of London. Series B: C., & Strassmann, J. E. (2005). The social parasite wasp Biological Sciences, 266(1415), 173–177. https://doi.org/10.1098/ Polistes atrimandibularis does not form host races. Journal rspb.1999.0618 of Evolutionary Biology, 18(5), 1362–1367. https://doi. Beibl, J., Stuart, R. J., Heinze, J., & Foitzik, S. (2005). Six origins of slavery org/10.1111/j.1420-9101.2005.00927.x in formicoxenine ants. Insectes Sociaux, 52(3), 291–297. https://doi. Field, J. (1992). Intraspecific parasitism as an alternative reproductive org/10.1007/s0004​0-005-0808-y tactic in nest-building wasps and bees. Biological Reviews, 67(1), 79– Beresford, J., Elias, M., Pluckrose, L., Sundström, L., Butlin, R. K., 126. https://doi.org/10.1111/j.1469-185X.1992.tb01659.x​ Pamilo, P., & Kulmuni, J. (2017). Widespread hybridization within Fitzpatrick, B. M., Fordyce, J. A., & Gavrilets, S. (2008). What, if anything, mound-building wood ants in Southern Finland results in cytonu- is sympatric speciation? Journal of Evolutionary Biology, 21(6), 1452– clear mismatches and potential for sex-specific hybrid breakdown. 1459. https://doi.org/10.1111/j.1420-9101.2008.01611.x Molecular Ecology, 26(15), 4013–4026. https://doi.org/10.1111/ Foitzik, S., DeHeer, C. J., Hunjan, D. N., & Herbers, J. M. (2001). mec.14183 Coevolution in host–parasite systems: Behavioural strategies of Bernasconi, C., Cherix, D., Seifert, B., & Pamilo, P. (2010). Molecular slave–making ants and their hosts. Proceedings of the Royal Society of of the Formica rufa group (red wood ants) (Hymenoptera: London. Series B: Biological Sciences, 268(1472), 1139–1146. https:// Formicidae): A new cryptic species in the Swiss Alps? Myrmecological doi.org/10.1098/rspb.2001.1627 News, 14, 37–47. Francœur, A. (1973). Révision taxonomique des espèces néarctiques du Bono, J. M., Blatrix, R., Antolin, M. F., & Herbers, J. M. (2007). Pirate groupe fusca, genre Formica (Formicidæ, Hymenoptera). Mémoires ants (Polyergus breviceps) and sympatric hosts (Formica occulta and de La Société Entomologique Du Québec, 3, 1–316. SAPP et al. | 2061

Funk, D. J. (2012). Of host forms and host races: Terminological issues in McCoy, K. D., Boulinier, T., Tirard, C., & Michalakis, Y. (2001). Host spec- ecological speciation. International Journal of Ecology, 2012, https:// ificity of a generalist parasite: Genetic evidence of sympatric host doi.org/10.1155/2012/506957 races in the seabird tick Ixodes uriae. Journal of Evolutionary Biology, Gardner, A., & West, S. A. (2004). Spite among siblings. Science, 14(3), 395–405. https://doi.org/10.1046/j.1420-9101.2001.00290.x 305(5689), 1413–1414. https://doi.org/10.1126/scien​ce.1103635 Mori, A., Grasso, D. A., Visicchio, R., & Le Moli, F. (2001). Comparison Gibbs, H. L., Sorenson, M. D., Marchetti, K., de L. Brooke, M., Davies, N. of reproductive strategies and raiding behaviour in facultative and B., & Nakamura, H. (2000). Genetic evidence for female host-specific obligatory slave-making ants: The case of Formica sanguinea and races of the common cuckoo. Nature, 407(6801), 183–186. https:// Polyergus rufescens. Insectes Sociaux, 48(4), 302–314. https://doi. doi.org/10.1038/35025058 org/10.1007/PL000​01782 Glasier, J. R., Acorn, J. H., Nielsen, S. E., & Proctor, H. (2013). Ants Pamilo, P., Gertsch, P., Thorén, P., & Seppä, P. (1997). Molecular population (Hymenoptera: Formicidae) of Alberta: A key to species based primar- genetics of social insects. Annual Review of Ecology and Systematics, ily on the worker caste. Canadian Journal of Identification, 28, 1–25. https://doi.org/10.1146/annur​ev.ecols​ys.28.1.1 22(22), 1–104. https://doi.org/10.3752/cjai.2013.22 Pritchard, J. K., Stephens, M., & Donnelly, P. (2000). Inference of pop- Goodloe, L. P., & Sanwald, R. (1985). Host specificity in colony-found- ulation structure using multilocus genotype data. Genetics, 155(2), ing by Polyergus lucidus queens (Hymenoptera: Formicidae). 945–959. https://doi.org/10.1111/j.1471-8286.2007.01758.x Psyche: A Journal of Entomology, 92(2–3), 297–302. https://doi. Queller, D. C. (1994). Genetic relatedness in viscous populations. org/10.1155/1985/69513 Evolutionary Ecology, 8(1), 70–73. https://doi.org/10.1007/BF012​ Goodloe, L. P., Sanwald, R., & Topoff, H. (1987). Host specificity in raiding 37667 behavior of the slave-making ant, Polyergus lucidus. Psyche: A Journal Read, A. F., & Taylor, L. H. (2001). The ecology of genetically diverse in- of Entomology, 94(1–2), 39–44. fections. Science, 292(5519), 1099–1102. https://doi.org/10.1126/ Griffin, A. S., & West, S. A. (2002). Kin selection: Fact and fiction. Trends science.1059410​ in Ecology and Evolution, 17(1), 15–21. https://doi.org/10.1016/S0169​ Rosenberg, N. A. (2004). distruct: A program for the graphical display of -5347(01)02355-2​ population structure. Molecular Ecology Notes, 4(1), 137–138. https:// Hamilton, W. D. (1964). The genetical evolution of social be- doi.org/10.1046/j.1471-8286.2003.00566.x haviour. I. Journal of Theoretical Biology, 7(1), 1–16. https://doi. Rothstein, S. I. (1990). A model system for coevolution: avian brood para- org/10.1016/0022-5193(64)90038​-4 sitism. Annual Review of Ecology and Systematics, 21, 481–508. Heinze, J., & Keller, L. (2000). Alternative reproductive strategies: A Sapp, J., Yost, J., & Lyon, B. (2020). Data from: Microsatellite evidence queen perspective in ants. Trends in Ecology & Evolution, 15(12), 508– shows host-associated population structure and relatedness among 512. https://doi.org/10.1016/S0169​-5347(00)01995​-9 neighbors in a social parasite (Polyergus mexicanus). v9. Dryad Dataset, Herbers, J. M. (2007). Watch your language! Racially loaded metaphors https://doi.org/10.5061/dryad.wstqjq2h9​ in scientific research. BioScience, 57(2)(2), 104–105. https://doi. Schär, S., Talavera, G., Espadaler, X., Rana, J. D., Andersen Andersen, A., org/10.1641/B570203 Cover, S. P., & Vila, R. (2018). Do Holarctic ant species exist? Trans- Hölldobler, B., & Wilson, E. O. (1990). The ants. Cambridge, MA: Harvard Beringian dispersal and homoplasy in the Formicidae. Journal of University Press. Biogeography, 45(8), 1917–1928. https://doi.org/10.1111/jbi.13380 Jaenike, J. (1981). Criteria for ascertaining the existence of host races. Schuelke, M. (2000). An economic method for the fluorescent labeling The American Naturalist, 117(5), 830–834. of PCR fragments. Nature Biotechnology, 18, 233–234. https://doi. Jakobsson, M., & Rosenberg, N. A. (2007). CLUMPP: A cluster match- org/10.1038/72708 ing and permutation program for dealing with label switching and Schumann, R. D., & Buschinger, A. (1994). Imprinting effects on host-se- multimodality in analysis of population structure. Bioinformatics, lection behaviour of colony-founding Chalepoxenus muellerianus 23(14), 1801–1806. https://doi.org/10.1093/bioinforma​ tics/​ (Finzi) females (Hymenoptera, Formicidae). Ethology, 97, 33–46. btm233 https://doi.org/10.1111/j.1439-0310.1994.tb01027.x​ Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, Schumann, R. D., & Buschinger, A. (1995). Imprinting effects on host-se- S., … Drummond, A. (2012). Geneious Basic: An integrated and ex- lection behavior of slave-raiding Chalepoxenus muellerianus (Finzi) tendable desktop software platform for the organization and analy- workers (Hymenoptera, Formicidae). Ethology, 99(3), 243–251. sis of sequence data. Bioinformatics, 28(12), 1647–1649. https://doi. https://doi.org/10.1111/j.1439-0310.1995.tb00898.x​ org/10.1093/bioin​forma​tics/bts199 Seifert, B. (2009). Cryptic species in ants (Hymenoptera: Formicidae) Konovalov, D. A., Manning, C., & Henshaw, M. T. (2004). kingroup: A pro- revisited: We need a change in the alpha-taxonomic approach. gram for Pedigree relationship reconstruction and Kin Group assign- Myrmecological News, 12, 149–166. ments using genetic markers. Molecular Ecology Notes, 4(4), 779–782. Seifert, B., Kulmuni, J., & Pamilo, P. (2010). Independent hybrid pop- https://doi.org/10.1111/j.1471-8286.2004.00796.x ulations of Formica polyctena x rufa wood ants (Hymenoptera: Liautard, C., & Keller, L. (2001). Restricted effective queen disper- Formicidae) abound under conditions of forest fragmentation. sal at a microgeographic scale in polygynous populations of the Evolutionary Ecology, 24(5), 1219–1237. https://doi.org/10.1007/ ant Formica exsecta. Evolution, 55(12), 2484–2492. https://doi. s10682-010-9371-8​ org/10.1111/j.0014-3820.2001.tb00763.x​ Sundström, L., Seppä, P., & Pamilo, P. (2005). Genetic population struc- Little, T. J., Watt, K., & Ebert, D. (2006). Parasite-host specificity: ture and dispersal patterns in Formica ants – A review. Annales Experimental studies on the basis of parasite adaptation. Evolution, Zoologici Fennici, 42(3), 163–177. https://doi.org/10.1007/BF001​ 60(1), 31–38. https://doi.org/10.1554/05-316.1 72934 Mackay, W., & Mackay, E. (2002). The ants of New Mexico (Hymenoptera: Thompson, J. N. (2005). Coevolution: The geographic mosaic of coevolu- Formicidae). Lewiston, NY: The Edwin Mellen Press. tionary arms races. Current Biology, 15(24), R992–R994. https://doi. Marchetti, K. (1998). Host-race formation in the Common Cuckoo. org/10.1016/j.cub.2005.11.046 Science, 282(1989), 471–472. https://doi.org/10.1126/scien​ Timms, R., & Read, A. F. (1999). What makes a specialist special? Trends in ce.282.5388.471 Ecology & Evolution, 14(9), 333–334. https://doi.org/10.1016/S0169​ Martin, S., & Drijfhout, F. (2009). A review of ant cuticular hydrocar- -5347(99)01697​-3 bons. Journal of Chemical Ecology, 35(10), 1151–1161. https://doi. Topoff, H. (1990). The evolution of slave-making behavior in the parasitic org/10.1007/s10886-009-9695-4​ ant genus Polyergus. Ethology Ecology & Evolution, 284–287. 2062 | SAPP et al.

Topoff, H., Lamon, B., Goodloe, L. P., & Goldstein, M. (1984). Social and exsecta. The American Naturalist, 186(6), 716–727. https://doi. orientation behavior of Polyergus breviceps during slave-making org/10.1086/683799 raids. Behavioral Ecology and Sociobiology, 15, 273–279. https://doi. West, S. A., Pen, I., & Griffin, A. S. (2002). Cooperation and competi- org/10.1007/BF00292989​ tion between relatives. Science, 296(5565), 72–75. https://doi. Torres, C. W., Tonione, M. A., Ramírez, S. R., Sapp, J. R., & Tsutsui, N. org/10.1126/science.1065507​ D. (2018). Genetic and chemical divergence among host races of Wisenden, B. D. (1999). Alloparental care in fishes. Reviews in Fish a socially parasitic ant. Ecology and Evolution, 8(23), 11385–11398. Biology and Fisheries, 9(1), 45–70. https://doi.org/10.1023/A:10088​ https://doi.org/10.1002/ece3.4547 65801329 Trager, J. C. (2013). Global revision of the dulotic ant genus Polyergus Zink, A. G. (2003). Intraspecific brood parasitism as a conditional repro- (Hymenoptera: Formicidae, Formicinae, Formicini). Zootaxa, 3722(4), ductive tactic in the treehopper Publilia concava. Behavioral Ecology 501–548. and Sociobiology, 54(4), 406–415. https://doi.org/10.1007/s0026​ Van Oosterhout, C., Hutchinson, W. F., Wills, D. P. M., & Shipley, P. 5-003-0649-5 (2004). micro-checker: Software for identifying and correcting ge- notyping errors in microsatellite data. Molecular Ecology Notes, 4(3), 535–538. https://doi.org/10.1111/j.1471-8286.2004.00684.x How to cite this article: Sapp JR, Yost J, Lyon BE. The socially Via, S. (2001). Sympatric speciation in : The ugly duckling parasitic ant Polyergus mexicanus has host-associated genetic grows up. Trends in Ecology & Evolution, 16(7), 381–390. https://doi. org/10.1016/S0169-5347(01)02188​ -7​ population structure and related neighbouring colonies. Mol Vitikainen, E. I. K., Haag-Liautard, C., & Sundström, L. (2015). Natal Ecol. 2020;29:2050–2062. https://doi.org/10.1111/mec.15468 dispersal, mating patterns, and inbreeding in the ant Formica